LegalizeDAG: make sure cast is unsigned before using FP_TO_UINT.

It's valid to use FP_TO_SINT when asking for a smaller type (e.g. all
"unsigned int16" values fit into a "signed int32"), but the reverse
isn't true.

Unfortunately, I'm not actually aware of any architecture with
asymmetric FP_TO_SINT and FP_TO_UINT handling and the logic happens to
work in the symmetric case, so I can't actually write a test for this.

llvm-svn: 210986

1 files changed, 4 insertions, 1 deletions

diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
index c76a52abe69..48eb86c49c3 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeDAG.cpp
@@ -2650,12 +2650,15 @@ SDValue SelectionDAGLegalize::PromoteLegalFP_TO_INT(SDValue LegalOp,
     NewOutTy = (MVT::SimpleValueType)(NewOutTy.getSimpleVT().SimpleTy+1);
     assert(NewOutTy.isInteger() && "Ran out of possibilities!");
 
+    // A larger signed type can hold all unsigned values of the requested type,
+    // so using FP_TO_SINT is valid
     if (TLI.isOperationLegalOrCustom(ISD::FP_TO_SINT, NewOutTy)) {
       OpToUse = ISD::FP_TO_SINT;
       break;
     }
 
-    if (TLI.isOperationLegalOrCustom(ISD::FP_TO_UINT, NewOutTy)) {
+    // However, if the value may be < 0.0, we *must* use some FP_TO_SINT.
+    if (!isSigned && TLI.isOperationLegalOrCustom(ISD::FP_TO_UINT, NewOutTy)) {
       OpToUse = ISD::FP_TO_UINT;
       break;
     }